Slider

• Slider 23 Oct 2019 22:59 UTC

  10 points

  in reply to: bendini's comment on: What are some un­pop­u­lar (non-nor­ma­tive) opinions that you hold?

  Fol­low­ing in­struc­tions doesn’t re­ally ring as a bell as a site goal. The set­ting of the ques­tion seems fair but the ill com­mit­ted in ig­nor­ing the con­text is differ­ent from di­s­obeyance.

• Slider 23 Oct 2019 13:50 UTC

  3 points

  on: Does the body have an al­most in­finite num­ber of po­ten­tial po­si­tions?

  I con­tam­i­nated my­self be­cause I thought it was a weird ques­tion by it­self and had a smell of the an­swer be­ing em­bed­ded in the con­text. In gen­eral “Does this anal­y­sis miss any­thing? I have not in­cluded anal­y­sis to avoid bias” isn’t re­ally an an­swer­able ques­tion.

  One of the tricks ot pon­der on mean­ings is to won­der what the mean­ing stands in op­po­si­tion to. To me it seems the state­ment plays the role of “enu­mer­at­ing the po­si­tions is not a fruit­ful way of ap­proach­ing han­dling of po­si­tions” and if it is not enu­mer­able then by the­o­rycraft­ing it is prob­a­bly some kind of in­finity. And the ex­am­ple “proof” seems to be that for any pro­posal enu­mer­at­ing scheme you can take a po­si­tion, ap­ply a small mu­ta­tor and end up with a po­si­tion that is not cov­ered by the origi­nal enu­mer­a­tion. A weaker ver­sions re­sults if only “”un­likely to be already enu­mer­ated”.

  Tak­ing the logic a bit fur­ther it seems that the prob­lem is already in “this po­si­tion” where the point­ing “this” can be am­bi­gious in micropa­ram­e­ters where try­ing to point to “one” po­si­tion ac­tu­ally re­sults in point­ing to a whole class of po­si­tions. Thus “count­ing” po­si­tions is prob­le­matic and not re­ally needed if we just want to un­der­tand po­si­tions and their pa­ram­e­ters/​po­si­tions. So rather than there be­ing a great many num­ber of them an anal­y­sis is em­ployed where we do not num­ber them.

  A similar thing could be made with wa­ter/​ice. If you have ice­cubes it might make sense to count them. How­ever if you have a cup of wa­ter or a block of ice it would be un­nat­u­ral to place a in­te­ger amount to de­scribe amount of wa­ter (al­thought with con­nec­tion to av­o­cadros num­ber there prob­a­bly is such a num­ber) but nei­ther it would make sense to say that a cup of wa­ter con­tains “in­finite” amount of wa­ter. One might still care about big cup of wa­ter vs small cup of wa­ter. How­ever the un­countable sense of size is not di­rectly re­lated to the countable sense of size (mea­sure one in liters and one in ob­jects).

• Slider 15 Oct 2019 17:05 UTC

  1 point

  in reply to: countedblessings's comment on: Sets and Functions

  You are mak­ing par­tially sense in that you are point­ing to a mod­el­ling style but it does leave me un­sure whether I can cor­rectly fill in the miss­ing bits. My think­ing is in­terferred a lot by “get­ter func­tions” that are of the form [code]func­tion get_color(self){ re­turn self.color }. One of the point of such is that at­tributes tend to be pri­vate but meth­ods are pub­lic and the pro­gram­mer should he need to do so could change the im­ple­men­ta­tion de­tails with­out mess­ing out­side cus­tomers. So the mod­eel­ing style shares a similar­ity that ob­jects are al­lowed to se­cretly have de­tails out­side of their in­ter­face. Sure if we have verbs and ob­jects mixed up but can ex­press ob­ject-like things as verbs by con­vert­ing ob­jects to verbs we only have to care about one ba­sic on­tol­ogy type. But I am un­sure whether I missed it or is it fort­com­ing why it is im­por­tant or valuable to fo­cus on the verbs.

  I am un­sure what rule-based­ness is but if it is differ­ent from ex­ten­sional con­cep­tion of func­tions then I would be su­per in­trigued. I can get that sens­ing should be mod­el­led with func­tions in that way but it seems con­tra­dic­tory how func­tions were sup­posed to be pre­dic­tion or evolu­tion mod­els. So if I have a(b(c(d))) does it mean that first d goes throught two kinds of evolu­tions and is then ob­served or d goes throught one kind of evolu­tion and then ob­ser­va­tion of that is ob­served. I am ex­pect­ing this kind of di­vi­son is not an ac­tual prob­lem but I can’t effortlessly go from the func­tion for­mal­ism to ob­ser­va­tion/​pre­dic­tion for­mal­isms and would likely make a lot of er­rors there.

• Slider 15 Oct 2019 0:05 UTC

  1 point

  in reply to: norswap's comment on: Sets and Functions

  That would the am­bush part?

• Slider 14 Oct 2019 23:43 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  I think in your treat­ment be­tween differ­ent monothe­is­tic gods there would be ev­i­dence that would favour one god over an­other even if both start similarly dys­mally low. How­ever if you had two gods and they are in­dis­t­in­guish­able from aliens then they should not be dis­t­in­guish­able from each other. I guess there are two senses in that in “ev­i­dence we cur­rently have” vs “ev­i­dence that could ever ex­ist”. Like I would think that god hy­poth­e­sis would not have in­creased prob­a­bil­ity for fly­ing saucers. But if god doesn’t raise the ex­pec­ta­tion of saucers does god raise the ex­pec­ta­tion of any­thing? If it doesn’t raise the ex­pec­ta­tion of any­thing then there is noth­ing to dis­agree about be­cause we don’t mean any­thing.

• Slider 14 Oct 2019 23:13 UTC

  2 points

  in reply to: valentinslepukhin's comment on: I would like to try dou­ble crux.

  What would count as suc­cess­ful con­tact? Can an out­side per­son ver­ify that con­tact has hap­pened? If some­one would con­vince you that your con­tact was ac­tu­ally con­fir­ma­tion bias would you change your opinion or prob­a­blity in god ex­ist­ing?

• Slider 14 Oct 2019 22:57 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  Para­phras­ing: By the mir­a­cle I un­der­stand event or se­ries of events that have ei­ther: a) ex­tremely low prob­a­bil­ity, b) seems to break the laws of Na­ture or c) have sig­nifi­cantly higher prob­a­bil­ity to oc­cur in the world with God rather than with­out.

  “Low prob­a­bil­ity events hap­pen all the time”

  ‘I do not be­lieve in “mir­a­cles”, in the sense that you prob­a­bly mean’

  The sense in which you replied to “mir­a­cles” seems not to be able to be un­der­stood in the literal defi­ni­tion pro­vided, so you ei­ther used your own pri­vate defi­ni­tion or did not be­lieve that their defi­ni­tion was ac­cu­rately spel­led out. Now the dis­cus­sion has shifted and there has been an ad­di­tonal fea­ture added that mir­a­cles are con­nected to chris­tian wor­ship. If you are as­sum­ing they have ad­di­tional prop­er­ties it might be fruit­ful to be ex­plicit about them.

• Slider 14 Oct 2019 22:43 UTC

  0 points

  in reply to: kithpendragon's comment on: I would like to try dou­ble crux.

  if the clair­voy­ant knows any thing ie their ex­pe­rience cor­re­lates to any­thing then the other be­ing is pre­sent. But I guess it would be hope­less to es­tab­lish how re­li­able the in­for­ma­tion chan­nel is us­ing that in­for­ma­tion chan­nel only.

• Slider 14 Oct 2019 17:19 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  It could be en­tirely con­sis­tent that the pro­gram­mer made na­ture_1 within the con­straints of na­ture_2 and “laws of na­ture” refer­ring to na­ture_2 laws is un­stan­dard and sur­pris­ing. it would be an as­sump­tion that na­ture_2 has time. Sure there are hard­ware con­straints but their ex­act shape is hard to ar­gue (I could come up with ex­am­ples how simu­lated time takes simu­la­tor space and not time).

• Slider 14 Oct 2019 16:58 UTC

  2 points

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  If the ev­i­dence would be ex­actly the same for god as for other things shouldn’t it be equally likely to be god rather than less likely?

• Slider 14 Oct 2019 16:56 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  Do you have rea­son to be­lieve that valentin­le­spukhin is in fact us­ing a differ­ent defi­ni­tion than he ex­plic­itly pro­vided here? The ex­plicit defi­ni­tion was of the form a,b or c and you claim to be­lieve that en­tities of type a ex­ist. Why would you not be­lieve that the dis­junc­tion ex­ists?

• Slider 14 Oct 2019 16:47 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  You are say­ing that you can to­tally com­mu­ni­cate with a non-ex­is­tent god so that point is only a sin­gle sided crux at the mo­ment.

  I don’t re­ally un­der­stand the differ­ence be­tween clair­voy­ant and tele­pathic. Either the con­tact mechanism is know or un­known. If it is known we can ar­gue what kind of en­tities can be in that kind of con­tact. If it is not known there is no point in differ­en­ti­at­ing be­tween differ­ent types as the de­tails could be any­thing.

• Slider 14 Oct 2019 12:17 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  Na­ture com­po­nent doesn’t re­ally break the anal­ogy and there are ways for which in can be made sense for part of the logic. If my pro­gram han­dles data a cer­tain way as a pro­gram­mer I would be free to al­ter that. I could for ex­am­ple go from us­ing sets to us­ing fuzzy sets. As I don’t want to po­si­tion my­self as a crux party I will re­frain from want­ing what om­nipo­tence should mean in this con­ver­sa­tion.

• Slider 14 Oct 2019 12:06 UTC

  1 point

  in reply to: valentinslepukhin's comment on: I would like to try dou­ble crux.

  3. What would you con­clude about ex­is­tence of god in the in­dis­t­in­guish­able case? Would the pro­gram­mer be god or would that be a sep­a­rate en­tity? Do you think the per­cep­tion is dis­t­in­guish­able?

• Slider 14 Oct 2019 11:57 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  If you would have tele­pathic con­tact with a for­eign in­tel­li­gence would that make you think that other com­mu­ni­ca­tion party ex­ists?

  What would con­vince you that an­other per­son beside your­self was hav­ing com­mu­ni­ca­tion con­tact with non-hu­man in­tel­li­gences?

  Do you lack the com­munca­tion po­ten­tial to non-hu­man in­tel­li­gences? If yes that would be ma­te­rial dis­agree­ment whether ev­ery per­son can.

• Slider 14 Oct 2019 11:44 UTC

  1 point

  in reply to: gilch's comment on: I would like to try dou­ble crux.

  Do you be­lieve in weak de­cay? Do you think weak de­cay is a low prob­a­blity event? Why does not weak de­cay fit within the sense of mir­a­cle here? Do you be­lieve ev­ery kind of chess game that has played out has played out mul­ti­ple times? If not why are the low fre­quency games not these low prob­a­blity event se­ries? What kinds of events would be raised in prob­a­blity if god would ex­ist? Would there be a sep­a­rat­ing sig­na­ture from aliens and other fan­tas­ti­cals?

• Slider 14 Oct 2019 11:34 UTC

  2 points

  in reply to: valentinslepukhin's comment on: I would like to try dou­ble crux.

  1. Are magic tricks mir­a­cles be­cause they seem to break the laws of Na­ture? Are lot­ter­ies mir­a­cles be­cause they are ex­tremely low prob­a­bil­ity? What counts as a law of na­ture? Does it fac­tor hu­man knowl­edge into it? Are su­perfluids an ex­am­ple of break­ing the law of na­ture of liquids?

• Slider 14 Oct 2019 11:17 UTC

  3 points

  in reply to: Evan Rysdam's comment on: What’s go­ing on with “prov­abil­ity”?

  What kind of refer­ence you are us­ing for your refer­ence to sets if not the ax­ioms? That reads to me as if “Are they just to­tally un­re­lated ob­jects to red busses that just hap­pen to be a bus and red?”

  Some times I have seen peo­ple ar­gue for ex­am­ple that the word “yel­low” is grounded by the set of all yel­low things. But usu­ally that kind of defi­ni­tion suffers from the list be­ing am­bi­gious/​in­suffi­cient. Like if a give a thing it ei­ther is or is not a mem­ber of that set. But list­ing all the mem­bers or oth­er­wise giv­ing some pro­ce­dure to give out all the mem­bers seems like is not the most nat­u­ral thing to do. Thus if you tried to take the carte­sian product of yel­low things and red things be­cause you can’t ex­em­plify a sam­ple just from the con­cept you can’t build up the product from mem­bers. The col­lec­tion of yel­low things prop­bably is not a set but it has many set-like prop­er­ties. By hav­ing a close in­ven­tory of sets prop­er­ties they can be dis­t­in­guished from con­fused or nearby no­tions.

  Another pos­si­ble imag­i­na­tion prompt would be a per­son faced with co­or­di­nates. Is there a real num­ber that you can spe­sify that the hu­man can point out on the x axis? No, they are always go­ing to be off. In the same way if you pre­sent the axis and ask the hu­man to point out their “favourite num­ber” (that is sup­posed to keep sta­ble) they will point out a slightly differ­ent real num­ber each time they sup­pos­edly point that point out. Such a per­son can’t provide a choice func­tion. It might still make sense to treat the per­son as be­ing able to spec­ify in­ter­vals, or re­fer to all of the points or be­ing able to refer­ence cross­ing points and oth­ers that have ge­o­met­ri­cal spe­sifi­ca­tion. But in gen­eral a line is not guaran­teed to have any refer­an­cale points fal­ling within it.

• Slider 13 Oct 2019 12:51 UTC

  2 points

  in reply to: elriggs's comment on: “Mild Hal­lu­ci­na­tion” Test

  It does not need to feel like con­text on the in­side and ar­guably if you are recog­nis­ing you are in a con­text you are think­ing about the situ­a­tion in a cer­tain situ­a­tion-in­de­pen­dent way.

  I don’t know if the ana­log hold but a typ­i­cal re­in­force­ment neu­ral net­work upon er­ror just back­prop­a­gates a weigth ad­just­ment. One could think that weights that are moved a lot are in­ter­pret to be “very in con­text” and weights that are moved a lilt­tle are “some­what out of con­text” which would lead a very fuzzy sense of con­text where there are no hard lines (well be­fore they are re­in­forced into place). While it might not be re­al­is­tic it would be com­pu­ta­tion­ally tractable to com­pare two neu­ral net­works which are more sharp or diffuse in the prop­a­ga­tion weight­ing which could lead to a differ­ent struc­ture in the high-fit state (or differ­ent times for reach­ing that high fit state).

  There are a lot of prej­u­dices so there tends to be hid­ing of these things if they are not strictly necce­sary. It feels good to be seen and have that cu­ri­os­ity and open­ness be a pos­i­tive in­ter­ac­tion.

• Slider 12 Oct 2019 14:38 UTC

  2 points

  in reply to: elriggs's comment on: “Mild Hal­lu­ci­na­tion” Test

  Well one of the other symp­toms sen­sory-over­load could be in­ter­preted as not do­ing the fil­ter­ing (I my­self don’t ex­hibit that so much but it is con­nected). In that way it is not strictly necce­sary. It’s also a mul­ti­stage pro­cess so you might have a global-lo­cal-global-lo­cal al­ter­a­tion on differ­ent parts of the hi­er­achy.

  It isn’t that ab­solute and while ev­ery­one prob­a­bly can man­age to fol­low the rules there might be a differ­ence how effort­ful it is. The the­ory might not be de­tailed enough to ad­dress ques­tions on that level and i don’t have the most up to date fa­mil­iar­ity with it (hav­ing wrong the­ory can do a lot of harm and it has fluxed quite a bit). While it is not con­text-blind­ness the re­lated trait of literal­mind­ed­ness would help with ex­plicit rules as you don’t have to “ap­ply com­mon sense” but just “ex­e­cute”. In a situ­a­tion where there are literal rules to be fol­lowed and con­text-sen­si­tive course of ac­tion con­text-blind­ness would drop the con­text sen­si­tive op­tion from be­ing rele­vant. [What I think was such a con­flict] (https://​​www.less­wrong.com/​​posts/​​G5TwJ9BGx­cgh5DsmQ/​​yes-re­quires-the-pos­si­bil­ity-of-no#oy­oNqpuaanWcXC4uG) a con­text heavy per­son might not even re­al­ise that a literal in­ter­pre­ta­tion was pos­si­ble.

  In a way jus­tice is sup­posed to be blind in a very near sense. If law is be­ing ap­plied to per­sons differ­ently it eas­ily and quickly be­comes un­fair. But if there is no spe­cial ad­here­ing to such prin­ci­ples the ap­pli­ca­tion tends to get un­even.